Method for tracking and evaluating nonparametric outline based on shape semanteme

Abstract

The invention relates to a method for tracking and evaluating a nonparametric outline based on shape semanteme. The method mainly comprises the following steps of: obtaining a corresponding relation between a nonparametric outline tracking result of the current frame of image and a standard outline position by adopting a minimum space distance method at first; performing a binary process on the nonparametric outline tracking result and a standard outline, and extracting a shape context characteristic of each outline point; and finally, searching by using a pyramid matching core from a partial region of an object relation obtained by the minimum space distance method so as to obtain a matching relation based on outline semanteme and distance measurement. By the method, evaluation of the effective nonparametric outline tracking result is realized; the method is universal and applicable to any outline tracking results; the outline expression is free from any limitation; a control point is not needed to express the outline and the outline is not needed to be closed; meanwhile, the searching efficiency is improved effectively, the precision of the evaluation of the nonparametric outline is enhanced, and the method has a good application prospect.

Description

technical field [0001] The present invention relates to the field of computer vision, in particular to an evaluation technology of nonparametric contour tracking based on shape semantics. Background technique [0002] Object tracking is a very important research problem in the field of computer vision, because it is the basis for the study of high-level vision problems such as motion analysis and action recognition. In terms of the representation of tracking results, the current tracking algorithms can be roughly divided into two categories: area tracking and contour tracking. [0003] Generally speaking, the area tracking algorithm uses the window form to represent the object. In most cases, the window adopts a rectangle or an ellipse. Since a rectangle or an ellipse can be represented by a series of parameters, the accuracy of the area tracking result can be obtained by simply calculating the distance between these parameters and the standard window parameters. The standa...

AI Technical Summary

Problems solved by technology

However, for non-parametric contour tracking methods, such as level set-based contour tracking, the length and shape of the contours are constantly changing, and there is no explicit expression, so the similarity between two contours is evaluated is very difficult

Images